Pentax K20D Report—Part IV: AF in Low-Light

By Carl Weese

Quite a few TOP readers have been asking about the low light AF performance of the Pentax K20D. It's no secret that the AF function in Pentax dSLR cameras through the K10D was not at the level of the industry leaders. It happens that I've been busy shooting all sorts of things in the six weeks or so I've had the camera, but low light situations didn't happen to be part of it. Then, this past Sunday, I went to a nearby Historical Society for the opening of an exhibit entitled "The Fabric of Marriage: Wedding Dresses." The title was irresistible.

Historical Society, Washington, Connecticut

The little town of Washington has a long history both as a retreat for the world-weary well-to-do, and as something of an artist colony. The combination continues to this day and makes for interesting exhibits at the Historical Society. This one included material reaching more than two centuries into the past, along with more recent material, displayed with a whimsy that approached the surreal.

Historical Society, Washington, Connecticut

Of course I started taking pictures the minute I walked in, then I realized I was testing the K20D in pretty difficult low-light conditions. The light was dim enough to need exposures in the range of 40th to 60th of a second at ƒ/4  or ƒ/3.5 using ISO 800. It was also really harsh with huge jumps from highlight to shadow. The photographer for the local paper took one look and got out a flash gun:

Historical Society, Washington, Connecticut

So how did the K20D do? I was once again impressed by the "unannounced" improvement in AF performance of the K20 over the K10. Given the situation, I chose to set the camera to S(ingle)AF, and Selective focus point. With this combination, you can use the control wheel to punch your way around the 13 AF points available in the finder. Focus is actuated when you press the shutter release half-way. If none of the focus points let you grab focus on the right part of the picture with exactly the framing you want, you can shift the framing before shooting. If you want to shoot another frame, you can grab focus again with much less disturbance to your framing than using center-only. In C(ontinuous)AF you are stuck with framings allowed by the need to shoot with an active focus point in the right place.

For, it turned out, 99 frames, the K20D with, for a while, the 35mm ƒ/2.8 Macro, and then the 21mm DA Limited, did at least as well as I could have managed with a first-rate manual film camera system. By which I mean a rangefinder Leica, or something like an old Nikon F3 with fast (ƒ/2 or ƒ/1.4) lenses. It flat-out did better than I could possibly expect to do by focusing this equipment manually. Between the relatively slow (dark finder) lenses and the inherent reduced perceptual size of the view in an APS-C sensor camera's finder, not to mention the hair-trigger abruptness of the focus helicoids in both these lenses, I wouldn't even try to focus visually.

At this light level, focus speed was not instantaneous. It wasn't "so fast I didn't notice." But I'd describe it as "fast." More important, it was positive. By that I mean that when I pressed the button halfway, the mechanism almost always made one direct move to focus, and stayed there. It only "hunted," a couple of times, and when it did hunt, I could immediately see that I'd selected a target that just didn't have enough detail or contrast for the system to grab. Even more important than that, when I examined the files on screen, it turned out that the AF had been consistently accurate.* You need all three of these, in some reasonable compromise combination. It doesn't matter how fast  and positive your AF system is when it grabs focus point if that point is a few inches behind or in front of the actual target. It doesn't matter how accurate the system is if the pictures all go away before you can release the shutter. You need a workable combination.

Historical Society, Washington, Connecticut

Next couple posts in the series will continue to look at issues revolving around AF, while looking at specifics of some lenses the Elves at Golden, Colorado, have sent me to review.

______________________

Carl

*AF performance and the K20 AF adjustment feature are proving to be an eye-opening experience that I will be reporting on further. Here's the teaser: tests on official focus targets at very close range, and flat targets at more realistic shooting distances, indicated one level of adjustment for my 21mm (the potential utility of the lens was the reason I bought into the system) but real-world results began to show front focus with this adjustment instead of the rear focus it had displayed before. Next time....

_______________________

Previously:
Pentax K20D Report Part I: Focus
Pentax K20D Report Part II: The CMOS Sensor
Pentax K20D Report Part III: Color, Tone: New Features

_______________________

Carl's website 
Working Pictures
WPII

Comments

[Michael]

Thanks for that, Carl (and Mike). Clear and useful, well illustrated. As usual.

[Bill Mitchell]

I'm beginning to think that I might have made a mistake by returning the K20D to B&H and getting a Canon XSi instead.

[Miserere]

Carl, many people on forums are reporting a big increase in AF performance with respect to the K10D. And what did Pentax say? Nothing. Well, actually, I believe they stated that the K20D had the same AF as the K10D. So which is it? I cannot believe so many people who have both the K10D and K20D can be wrong about the improved AF, but if it *is* better, then Pentax must have improved it somehow.

Or maybe, we're witnessing the first case of natural selection in the technological wild. I random binary mutation has improved the K20D's AF, which will result in an improved ability to procreate and produce faster focusing offspring.

OK, so I need to iron out some details of my theory...

[Scott Kirkpatrick]

Carl, 1/40 at f/4 and ISO800 doesn't sound all that dark. How about 1/10 @ f/2.8 and ISO 800? Is everything still working crisply?

scott

[switters]

I'm mildly curious about this, since I actually sold my K10D about six months ago because of the slow AF in low light. Unfortunately, it was also slow with the Limited primes (43 & 77 in particular, less so with the 21, 31 & 70) even in good light. Since those primes were the main reason I chose a Pentax DSLR, it didn't make sense to stick with the system.

I've missed those lenses, however, and if the K20D AF really is closer to Canon and Nikon's AF I'd be interested. Very interested.

[xtoph]

from above (scott) "Carl, 1/40 at f/4 and ISO800 doesn't sound all that dark."

my sentiments exactly. on top of which, you have high contrast subjects (good af targets) under spot lighting. while the contrasty light is obviously challenging for exposure, it is the opposite of challenging for af.

i routinely shoot in much, much lower light (more like iso1600, f/1.4, 1/25sec). my camera system often focusses even in light that low, provided i can point it at a contrasty target (which i can't always). moreover, i could focus an f3 and a leica rf in light that low or lower reliably and quickly. it seems disingenuous to me to compare your situation to those systems here.

i wish you the best with the pentax. it has features i'd like on my camera, and generally seems like a great kit. obviously, it suits you. but i'm still waiting to see an evaluation of af in challenging conditions.

[Carl Weese]

Scott,

Given the relative rarity of fast (like the old days with film cameras) lenses today and the relatively good quality of results at ISO 800, plus the usefulness of various anti-shake systems, I agree that 1/10th @ f/2.8, ISO 800 is an interesting test point.

I just located a situation around the house that met these conditions and shot some tests with the 35mm f/2.8 macro from a distance of eight or ten feet. Focus was slower, consistently taking about half a second to settle. The shots were correctly focused on the target point. I'll be interested to compare that performance to competitive equipment.

[Luke]

I upgraded from a K10D to the K20D. And indeed the autofocus seems like it was redesigned completely.

From what I've read, it seems the early K10D units were plagued with AF issues. Pentax fixed the problem in later units. So if you bought a K10D in the latter part of its production run, you're not supposed to see an improvement in autofocus performance between the K10D and K20D.

Those of us who bought the K10D early on do however see a huge improvement. So much so that I won't use my K10D anymore.

[Sune]

Thanks for your work, Carl. Always a pleasure reading the site

[Adam Maas]

I had some issues with my late K10D, but low light for me was 1/8 second at ISO1600 and f2.8

Now I'm shooting at 1/8-1/15 at ISO 6400 and f3.5-f4 on my D300 and AF is much better.

As to the K10D/K20D AF differences, they share an AF system, but not necessarily AF firmware. Likely all updates were in the firmware. There's likely a lot more performance to be found in the AF unit given its similarities to Nikon's Multi-CAM2000 unit from the D2's and F6, which is very similar from a hardware point of view, but much higher performance.